April 12, 2026
Yesterday I wrote 88% of a production codebase using AI agents I built. And I still feel like an idiot.
I spend every waking hour in AI. Building production systems on Claude, running MCP servers, orchestrating multiple LLMs, shipping RAG pipelines that serve thousands of requests. Reading every paper, every release, every benchmark. But you know what? The model I mastered last week is already obsolete. The framework I learned yesterday has three better alternatives today. The pattern I shipped three months ago didn't exist when I started building it.
So if I can't keep up, how are you supposed to?
You're that CTO with 30 minutes a week to research AI. You're trying to squeeze in a ChatGPT experiment between board meetings and budget reviews. Or you're that health system exec who got told to "develop an AI strategy" by someone who thinks AI means replacing doctors with chatbots. Maybe you're the PE partner evaluating AI-native portfolio companies while your inbox fills with pitches about "revolutionary" tools that'll be dead in six months.
You feel lost because you are lost. We all are.
The people who seem confident right now? They're lying, selling you something, or operating in such narrow lanes they've convinced themselves their slice represents the whole pie. The vendor promising their platform solves everything? They know it doesn't. The consultant with the 47-slide deck on "AI transformation"? They're reading the same blog posts you are, just charging more for it.
Nobody knows what they're doing. Not the researchers publishing papers, not the VCs writing checks, not the engineers building products. We're all making it up as we go.
And that's actually fine.
Because the only real advantage right now isn't expertise. It's honesty about what you don't know. It's building systems that can survive being wrong. It's admitting to your team that you're learning alongside them instead of pretending you've got it figured out.
You want to know why governance matters more than expertise right now? Because governance is just structured humility. It's admitting you need guardrails when you can't see around the next corner. Real governance isn't compliance theater or checkbox exercises. It's acknowledging you're building on shifting sand and planning accordingly.
When I set up those AI agents yesterday, I didn't build them to be perfect. I built them to be replaceable. Because next week there'll be a better model, a cleaner approach, a tool that makes today's cutting-edge look primitive. The code they wrote will outlive the agents that wrote it.
So stop trying to become an AI expert. Start becoming comfortable with not knowing. Build for change, not permanence. Hire people who can learn fast, not people who claim they've already learned everything. Make decisions you can reverse when you're wrong, because you will be wrong.
That CTO with 30 minutes? Use it to understand what questions to ask, not to find answers that'll expire before next quarter. That health exec? Focus on problems that need solving regardless of which AI wins. That PE partner? Invest in teams that can pivot, not products that can't.
We're all going to look back at what we're building today and laugh. Or cry. Probably both. But the ones who'll survive are the ones who knew they were guessing and built accordingly.